The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the Bash(*) tool to perform system-level operations, including symlinking files, running latexmk/pdflatex for compilation, and executing Python scripts. This broad access is necessary for its function but increases the impact of potential command injection if input parameters are not properly sanitized.
[DYNAMIC_EXECUTION]: In Phase 7, the skill dynamically generates a Python script (slides/generate_pptx.py) and executes it using python3. While intended to facilitate PowerPoint export, runtime script generation and execution is a high-risk pattern that can be exploited to run arbitrary code if the generation logic is influenced by malicious input data.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection as it ingests untrusted content from LaTeX files (paper/sections/*.tex) and potentially external style references.
Ingestion points: Reads local .tex files and external sources via the --style-ref argument.
Boundary markers: The instructions do not specify the use of clear delimiters or 'ignore' instructions when processing this content.
Capability inventory: The skill has Bash(*) access, file write capabilities, and the ability to execute generated Python code.
Sanitization: There is no mention of escaping or sanitizing the content extracted from the paper before it is used to generate slides or scripts.
[EXTERNAL_DOWNLOADS]: The STYLE_HELPER logic (extract_paper_style.py) accepts HTTP(S) URLs and arXiv IDs as sources. This results in the fetching of external, potentially untrusted content into the skill's processing pipeline.

paper-slides